EP0760574A2 - Echokompensatorsystem mit gemeinschaftlichem Koeffizientsspeicher - Google Patents

Echokompensatorsystem mit gemeinschaftlichem Koeffizientsspeicher Download PDF

Info

Publication number
EP0760574A2
EP0760574A2 EP96305918A EP96305918A EP0760574A2 EP 0760574 A2 EP0760574 A2 EP 0760574A2 EP 96305918 A EP96305918 A EP 96305918A EP 96305918 A EP96305918 A EP 96305918A EP 0760574 A2 EP0760574 A2 EP 0760574A2
Authority
EP
European Patent Office
Prior art keywords
filter
coefficient memory
background
echo
echo canceller
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP96305918A
Other languages
English (en)
French (fr)
Inventor
Vasu Iyengar
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
AT&T Corp
Original Assignee
AT&T Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by AT&T Corp filed Critical AT&T Corp
Publication of EP0760574A2 publication Critical patent/EP0760574A2/de
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04M—TELEPHONIC COMMUNICATION
    • H04M9/00—Arrangements for interconnection not involving centralised switching
    • H04M9/08—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic
    • H04M9/082—Two-way loud-speaking telephone systems with means for conditioning the signal, e.g. for suppressing echoes for one or both directions of traffic using echo cancellers

Definitions

  • the invention relates generally to echo cancellers systems, and more particularly to echo canceller systems with at least two echo cancellers.
  • Speech typically results in reflected waves.
  • the reflected wave arrives a very short time after the direct sound, it is perceived as a spectral distortion or reverberation.
  • the reflection arrives a few tens of milliseconds after the direct sound, it is heard as a distinct echo.
  • Such echoes may be annoying, and under extreme conditions can completely disrupt a conversation.
  • Line echoes i.e., electrical echoes
  • the hybrid transformer passes the far-end signal at the four-wire receive port through to the two-wire transmit port without allowing leakage into the four-wire transmit port.
  • this would require exact knowledge of the impedance seen at the two-wire ports, which in practice varies widely and can only be estimated.
  • the leaking signal returns to the far-end talker as an echo.
  • the situation can be further complicated by the presence of two-wire toll switches, allowing intermediate four-two-four wire conversions internal to the network. In telephone connections using satellite links with round-trip delays on the order of 600 ms, line echoes can become particularly disruptive.
  • Acoustic echoes occur in telecommunications networks due to acoustic coupling between a loudspeaker and a microphone (e.g., in a speakerphone).
  • a microphone e.g., in a speakerphone
  • an acoustic reflection of the far-end talker through the near-end conference room is returned to the far-end talker as an echo.
  • Acoustic echo cancellation tends to be more difficult than line echo cancellation since the duration of the acoustic path is usually several times longer (100-400 ms) than typical electrical line paths (20 ms), and the acoustic path may change rapidly at any time due to opening doors, moving persons, changing temperatures, etc.
  • Echo suppressors have been developed to control line echoes in telecommunications networks. Echo suppressors decouple the four-wire transmit port when signal detectors determine that there is a far-end signal at the four-wire receive port without any near-end signal at the two-wire receive port. Echo suppressors, however, are generally ineffective during double-talking when speakers at both ends are talking simultaneously. During double-talk, the four-wire transmit port carries both the near-end signal and the far-end echo signal. Furthermore, echo suppressors tend to produce speech clipping, especially during long delays caused by satellite links.
  • Echo cancellers have been developed to overcome the shortcomings of echo suppressors.
  • Echo cancellers include an adaptive filter and a subtracter.
  • the adaptive filter attempts to model the echo path.
  • the incoming signal is applied to the adaptive filter which generates a replica signal.
  • the replica signal and the echo signal are applied to the subtracter.
  • the subtracter subtracts the replica signal from the echo signal to produce an error signal.
  • the error signal is fed back to the adaptive filter, which adjusts its filter coefficients (or taps) in order to minimize the error signal.
  • the filter coefficients are stored in a coefficient memory. In this manner, the filter coefficients converge toward values that optimize the replica signal in order to cancel the echo signal. Echo cancellers offer the advantage of not disrupting the signal path.
  • Echo cancellers were first deployed in the U.S. telephone network in 1979, and currently are virtually ubiquitous in long-distance telephone circuits. See generally Messerschmitt, "Echo Cancellation in Speech and Data Transmission", IEEE Journal on Selected Areas in Communications, Vol. SAC-2, No. 2, March 1984, pp. 283-298; and Tao et al., "A Cascadable VLSI Echo Canceller", IEEE Journal on Selected Areas in Communications, Vol. SAC-2, No. 2, March 1984, pp. 298-303.
  • Double-talk detectors are commonly used for disabling the adaptation as double-talking occurs. Double-talk detectors may employ, for instance, Geigel's test to detect double-talking.
  • double-talk detectors fail to indicate the presence of double-talking for a time period (e.g., a whole syllable) after double-talking begins. During this time period, the coefficients may drift and lead to howling as mentioned above. Furthermore, double-talking becomes increasingly difficult to detect as the acoustic echo becomes large in comparison to the near-end signal.
  • Ochiai et al. discloses an echo canceller with a non-adaptive filter and an adaptive filter arranged in parallel.
  • the adaptive filter includes an adaptation processor, whereas as non-adaptive filter is devoid of an adaptation processor and must be updated by the adaptive filter.
  • Each filter generates a replica of the echo signal.
  • the filter coefficients of the adaptive filter are transferred into the non-adaptive filter when the replica signal from the adaptive filter provides a better estimate of the echo signal than the replica signal of the non-adaptive filter. Therefore, during uncorrelated double-talking, the non-adaptive filter is relatively immune from coefficient drift in the adaptive filter.
  • the echo canceller in Yatsuzuka et al. includes an adaptive main echo estimator and an adaptive sub echo estimator arranged in cascade.
  • the main echo estimator and the sub echo estimator each include adaptation processors.
  • the main echo estimator generates a replica of the echo signal, whereas the sub echo estimator generates a replica of a send output signal.
  • the main echo estimator has a small step gain for updating filter coefficients so that the response to an echo is slow, while the sub echo estimator has a large step gain and quick response for an echo.
  • the two echo estimators operate independently from each other.
  • the filter coefficients in the sub echo estimator are accumulated on related filter coefficients in the main echo estimator.
  • a reset mode the coefficients of the sub echo estimator are reset, and then the mode is switched to the ordinary mode.
  • Echo canceller systems which include first and second echo cancellers for cancelling respective first and second echos across different echo paths.
  • an acoustic echo path may arise between a loudspeaker and a microphone
  • a line echo path may arise at a hybrid transformer which connects the set's four-wire system to a two-wire local customer loop.
  • a first echo canceller may be used for cancelling the line echo
  • a second echo canceller may be used for cancelling the acoustic echo.
  • Each echo canceller is likely to require continuous adaptive processing, which places heavy demands on processing power (or MIPs).
  • first and second echo cancellers may each include two or more filters, as described by Ochiai et al.., and Yatsuzuka et al., supra, with each filter requiring its own coefficient memory (or memory space).
  • echo canceller systems may require a large amount of processing power and memory space.
  • a primary aspect of the invention is an echo canceller system with a shared coefficient memory.
  • an echo canceller system comprises first and second echo cancellers.
  • Each echo canceller includes a foreground filter and a background filter.
  • the foreground filters have dedicated coefficient memories, whereas the background filters do not.
  • the echo canceller system also includes a shared coefficient memory, and a controller for switching the shared coefficient memory between the background filters. The switching includes resetting the shared coefficient memory to prevent any transfer of filter coefficients between the background filters.
  • the background filters operate one at a time, depending on which is coupled to the shared coefficient memory. Therefore, the background filters share memory space and processing power.
  • each echo canceller updates the foreground filter, and the foreground filter provides the actual echo cancellation.
  • the foreground filters operate simultaneously, although the foreground filters are updated one at a time, depending on which background filter is coupled to the shared coefficient memory. Thus, coefficient updates for the foreground filters are initially stored in the shared coefficient memory.
  • the echo canceller system includes a receive path with a receive input port and a receive output port, and a send path with a send input port and a send output port, with the first echo canceller cancelling an echo between the receive output port and the send input port, and the second echo canceller cancelling an echo between the send output port and the receive input port.
  • the controller switches the shared coefficient memory from the second background filter to the first background filter in response to signals at the receive path and the send path indicating a first state, and the controller switches the shared coefficient memory from the first background filter to the second background filter in response to signals at the receive path and the send path indicating a second state. In this manner, the controller couples the coefficient memory to the echo canceller with the greatest need for adaptive operation.
  • the invention is especially well-suited for use in a loudspeaking telephone set, in which the first echo canceller cancels a line echo through a hybrid transformer in response to a near-end signal without a far-end signal (i.e., transmit state), and the second echo canceller cancels an acoustic echo between a loudspeaker and a microphone in response to a far-end signal without a near-end signal (i.e., receive state).
  • the first and second echo cancellers provide simultaneous echo cancellation from the foreground filters, although the echo cancellers are updated one at a time, depending on whether there is a transmit state or a receive state.
  • FIG. 1 shows a simplified schematic diagram illustrating the manner in which an echo canceller is generally used as part of a loudspeaking telephone set.
  • Echo canceller 10 includes a receive path 12 with receive input port 14 and receive output port 16, and send path 18 with send input port 20 and send output port 22.
  • a receive input signal x(t), representing a far-end signal, is applied to receive input port 14.
  • Receive input signal x(t) is coupled via path 14 to receive output port 16, which couples signal x(t) to loudspeaker 24.
  • Acoustic echo path 26 with transfer characteristic H[t] is located between loudspeaker 24 and microphone 28. Echo path 26 causes signal x(t) at loudspeaker 24 to appear as echo signal H[x(t)] at microphone 28.
  • Microphone 28 also generates near-end signal u(t) due to speech at the near-end. Therefore, microphone 28 generates send input signal s(t) consisting of near-end signal u(t) added to echo signal H[x(t)] by way of superposition.
  • Echo canceller 10 further comprises adaptive digital transversal filter 30 and subtracter 32.
  • Adaptive filter 30 has a finite impulse response (FIR), and its filter coefficients are adaptively updated to model the transfer characteristic H[t] at sample intervals.
  • Adaptive filter 30 synthesizes a replica signal y(t) as an estimate of the undesired echo signal H[x(t)] in response to receive input signal x(t).
  • FIR finite impulse response
  • Subtracter 32 subtracts replica signal y(t) from send input signal s(t) to form error signal e(t).
  • Error signal e(t) is coupled to send output port 22 to provide the send output signal.
  • Error signal e(t) is also fed back to adaptive filter 30.
  • the transfer characteristic H[t] of echo path 26 will be time-varying.
  • Echo signal H[x(t)] approximates the linear convolution of signal x(t) with the impulse response h(t) of echo path 26. Therefore, adaptive filter 30 adjusts its impulse response w(t) as best it can to match impulse response h(t).
  • the adaptive adjustment of filter 32 is controlled by error signal e(t). This adaptive adjustment is continued as long as there is a correlation between error signal e(t) and receive input signal x(t).
  • adaptive filter 30 When receive input signal x(t) is present and near-end signal u(t) is absent (i.e., far-end speech without near-end speech), adaptive filter 30 generates replica signal y(t) as a reliable estimate of echo signal H[x(t)]. When, however, both x(t) and u(t) are present (i.e., double-talking), adaptive filter 30 can become grossly misadjusted due to near-end signal u(t) as a disturbing factor in error signal e(t). This misadjustment prevents replica signal y(t) from providing a reliable estimate of H[x(t)], in which case H[x(t)] is improperly or inadequately canceled.
  • Discrete-time modeling can be obtained by assuming in FIG. 1 that signals x(t) and s(t) are applied to analog-to-digital converters before being applied to input ports 14 and 20, respectively, and likewise that signals x(t) and e(t) are received from digital-to-analog converters coupled to output ports 16 and 22, respectively, and further that all the relevant signals in echo canceller 10 are digital signals.
  • Digital-to-analog and analog-to-digital converters may be employed for connecting the echo canceller 10 with analog channels.
  • FIG. 2 shows a schematic diagram of an embodiment of an echo canceller system according to the invention.
  • the echo canceller system is part of an integrated circuit chip, which is part of a loudspeaking telephone set 100.
  • Echo canceller system 110 includes a receive path 112 with receive input port 114 and receive output port 116, and send path 118 with send input port 120 and send output port 122.
  • Port 116 is coupled via digital-to-analog converter 124 to loudspeaker 126.
  • Microphone 128 is coupled via analog-to-digital converter 130 to port 120.
  • Acoustic echo path 132 is disposed between loudspeaker 126 and microphone 128.
  • Port 122 is coupled via digital-to-analog converter 134 to port 136 of hybrid transformer 138, and port 140 of hybrid transformer 138 is coupled via analog-to-digital converter 142 to port 114.
  • Hybrid transformer 138 provides conversion between two-wire signals and four-wire signals.
  • Hybrid transformer ports 136 and 140 interface with a four-wire full-duplex system inside telephone set 100, whereas hybrid transformer ports 146 and 148 interface with a two-wire local customer loop outside telephone set 100.
  • Hybrid transformer 138 also provides line echo path 144 between ports 122 and 114.
  • Echo canceller system 110 also includes first and second echo cancellers.
  • the first echo canceller includes first foreground filter 152 coupled to first foreground subtracter 154, first adaptive background filter 156 coupled to first background subtracter 158, and first control unit 160.
  • the second echo canceller includes second foreground filter 162 coupled to second foreground subtracter 164, second adaptive background filter 166 coupled to second background subtracter 168, and second control unit 170.
  • the first echo canceller is designed to cancel a line echo across line echo path 144.
  • the second echo canceller is designed to cancel an acoustic echo across acoustic echo path 132.
  • foreground filter 152 provides the actual echo cancellation
  • control unit 160 determines when filter coefficients of background filter 156 can be used to update filter coefficients of foreground filter 152 in accordance with a first function.
  • foreground filter 162 provides the actual echo cancellation
  • control unit 170 determines when filter coefficients of background filter 166 can be used to update filter coefficients of foreground filter 162 in accordance with a second function.
  • foreground filters 152 and 162 may be adaptive filters, in accordance with Yatsuzuka et al., supra, or, preferably, non-adaptive filters.
  • the update functions include accumulating the background filter coefficients on the foreground filter coefficients.
  • first subtracters 154 and 158 and second subtracters 164 and 168 are shown as arranged in cascade, alternatively, the first subtracters 154 and 158 can be arranged in parallel, and likewise, the second subtracters 164 and 168 can be arranged in parallel, in accordance with Ochiai et al., supra. With the parallel arrangement, the update functions include transferring the background filter coefficients to the foreground filter coefficients.
  • Foreground filters 152 and 162 each have their own dedicated coefficient memories or memory space. For instance, a first coefficient memory (not shown) is used exclusively by filter 152, and a second coefficient memory (not shown) is used exclusively by filter 162. Background filters 156 and 166, however, lack dedicated coefficient memories. Accordingly, echo canceller system 110 further includes controller 172 and shared coefficient memory 174. Controller 172 switches shared coefficient memory 174 between background filters 156 and 166 so that one but not both of background filters 156 and 166 can store its respective filter coefficients in shared coefficient memory 174. Thus, controller 174 prevents background filters 156 and 166 from simultaneously accessing shared coefficient memory 174.
  • controller 174 resets shared coefficient memory 174 when shared coefficient memory 174 is switched between background filters 156 and 166, thereby preventing any transfer of filter coefficients between background filters 156 and 166.
  • shared coefficient memory 174 need not simultaneously store filter coefficients associated with background filters 156 and 166.
  • foreground filters 152 and 162 operate simultaneously, background filters 156 and 166 operate one at a time, depending on which background filter has access to shared coefficient memory 174. Consequently, control units 160 and 170 perform their respective update functions one at a time, depending on which background filter is coupled to the shared coefficient memory.
  • controller 172 performs the switching operation in response to signal levels at receive path 112 and send path 118.
  • Shared coefficient memory 174 is coupled to background filter 156 and decoupled from background filter 166, thereby allowing background filter 156 to operate and preventing background filter 166 from operating, when the signal level at send path 118 exceeds the signal level at receive path 112.
  • shared coefficient memory 174 is coupled to background filter 166 and decoupled from background filter 156, thereby allowing background filter 166 to operate and preventing background filter 156 from operating, when the signal level at receive path 112 exceeds the signal level at send path 118.
  • echo canceller system 110 For purposes of illustration, the operation of echo canceller system 110 is described with reference to a transmit state and a receive state.
  • the signals associated with the transmit state are shown inside arrow symbols ⁇ >, and the signals associated with the receive state are shown inside bracket symbols ⁇ ⁇ .
  • Detectors such as state controllers for determining which talker is active in a voice-switched speakerphone system are well known in the art. Such detectors generally switch states in response to signal levels at the send and receive paths. Many switching schemes are available. See, for instance, Sondhi et al., "Silencing Echos on the Telephone Network", Proceedings of the IEEE, Vol. 68, No. 8, August 1980, pp. 948-963; Clemency et al., "Functional Design of a Voice-Switched Speakerphone", The Bell System Technical Journal, Vol. XL, No. 3, May 1961, pp.
  • the transmit state may occur when send input signal u(k) is above a first threshold, and receive input signal x(k) is below a second threshold.
  • the receive state may occur when receive input signal x(k) is above a third threshold, and send input signal u(k) is below a fourth threshold.
  • the first, second, third and fourth thresholds may be essentially identical.
  • the thresholds may relate, for instance, to the instantaneous signal powers of signals x(k) and u(k) averaged over a specific period of time.
  • send input signal u(k), representing a near-end signal generated by a near-end source is present at send input port 120
  • receive input signal x(k), representing a far-end signal generated by a far-end source is absent at receive input port 114.
  • Line echo path 144 has a transfer characteristic of P[z].
  • the receive input signal consists of line echo signal P[u(k)].
  • controller 172 couples shared coefficient memory 174 to background filter 156, decouples shared coefficient memory 174 from background filter 166, and resets shared coefficient memory 174 thereby erasing the filter coefficients for background filter 166.
  • Foreground filter 152 continues to store first foreground filter coefficients in a first coefficient memory, and generates a replica signal y 1F (k), as an estimate of echo signal P[u(k)], in response to signal u(k).
  • echo signal P[u(k)] is applied to the addend input port, and replica signal y 1F (k) is applied to the subtrahend input port.
  • Subtracter 154 generates error signal e 1F (k), representing the difference between signals P[u(k)] and y 1 (k). Error signal e 1F (k) is transferred along receive path 112 to receive output port 116.
  • background filter 156 stores first background filter coefficients in shared coefficient memory 174, and generates a replica signal y 1B (k), as an estimate of signal y 1F (k), in response to signal u(k).
  • signal y 1F (k) is applied to the addend input port, and replica signal y 1B (k) is applied to the subtrahend input port.
  • Subtracter 158 generates error signal e 1B (k), representing the difference between signals e 1F (k) and y 1F (k). Error signal e 1B (k) is applied to a feedback input port of background filter 156 for further adaptive processing.
  • Control unit 160 measures error signals e 1F (k) and e 1B (k) to determine whether foreground filter 152 should be updated.
  • Background filter 166 is rendered inoperative by being decoupled from shared coefficient memory 174, although foreground filter 162 continues to store its filter coefficients in the second coefficient memory and operate during the transmit state. For convenience of explanation, it is assumed that any acoustic echo signal H[e 1F (k)] that occurs at port 120 is negligible.
  • controller 172 couples shared coefficient memory 174 to background filter 166, decouples shared coefficient memory 174 from background filter 156, and resets shared coefficient memory 174 thereby erasing the filter coefficients for background filter 156.
  • Foreground filter 162 stores second foreground filter coefficients in a second coefficient memory, and generates a second replica signal y 2F (k), as an estimate of echo signal H[x(k)], in response to signal x(k).
  • Subtracter 164 receives signal H[x(k)] at an addend input port, and receives signal y 2F (k) at a subtrahend input port.
  • Subtracter 164 generates error signal e 2F (k), representing the difference between signals H[x(k)] and y 2F (k). Error signal e 2F (k) is transferred along send path 118 to send output port 122.
  • background filter 166 stores second background filter coefficients in shared coefficient memory 174, and generates a replica signal y 1B (k), as an estimate of signal e 2F (k), in response to signal u(k).
  • signal e 2F (k) is applied to the addend input port, and replica signal y 1B (k) is applied to the subtrahend input port.
  • Subtracter 168 generates error signal e 1B (k), representing the difference between signals e 2F (k) and y 2F (k). Error signal e 1B (k) is applied to a feedback input port of background filter 166 for further adaptive processing.
  • Control unit 170 measures error signals e 2F (k) and e 2B (k) to determine whether foreground filter 162 should be updated.
  • Background filter 156 is rendered inoperative by being decoupled from shared coefficient memory 174, although foreground filter 152 continues store its filter coefficients in the first coefficient memory and operate during the receive state. For convenience of explanation, it is assumed that any line echo signal P[e 2F (k)] that occurs at port 114 is negligible.
  • FIG. 3 shows a schematic diagram of an embodiment of a portion of echo canceller system 110.
  • background filters 156 and 166, controller 172, and shared coefficient memory 174 are implemented by U register 202, control unit 204, convolution circuit 206, coefficient register 208, adaptation processor 210, and X register 212.
  • Signal paths 112, 118, 180, 182, 184, 186, 188 and 190 are shown for purposes of clarity.
  • Registers 202 and 212 are recirculated shift registers for storing samples of signals u(k) and x(k), respectively.
  • Control unit 204 includes a signal detector (not shown) for detecting the levels of signals u(k) and x(k).
  • Coefficient register 208 may be implemented, for instance, by random access memory (RAM).
  • Control unit 204 allows background filter 156 to operate by coupling U register 202 to convolution circuit 206 and to adaptation processor 210, signal paths 118 and 180 to adaptation processor 210, signal path 182 to convolution circuit 206, and signal path 184 to coefficient register 208, and decoupling the X register and signal paths 186, 188 and 190.
  • control unit allows background filter 166 to operate by coupling X register 212 to convolution circuit 206 and to adaptation processor 210, signal paths 112 and 186 to adaptation processor 210, signal path 188 to convolution circuit 206, and signal path 190 to coefficient register 208, and decoupling the U register and signal paths 180, 182 and 184.
  • convolution circuit 204 During operation of background filter 156, convolution circuit 204 generates replica signal y 1B (k) by performing linear convolution on the samples stored in U register 202 with the filter coefficients stored in coefficient register 208.
  • adaptation processor 210 adaptively corrects the filter coefficients stored in coefficient register 208, in response to signals u(k) and e 1B (k), in order to minimize the level of signal e 1B (k).
  • convolution circuit 204 generates replica signal y 2B (k) by performing linear convolution on the samples stored in X register 212 with the filter coefficients stored in coefficient register 208.
  • adaptation processor 216 adaptively corrects the filter coefficients stored in coefficient register 208, in response to signals x(k) and e 2B (k), in order to minimize the level of signal e 2B (k).
  • the memory space of coefficient register 208, and the processing power of adaptation processor 210 are shared between background filters 156 and 166.
  • Adaptation processor 210 may employ various adaptation algorithms that are well known in the art.
  • LMS Least Mean Square
  • NLMS Normalized Least Mean Square
  • LS Least Squares
  • U register 202 stores as many samples of u(k) as are necessary for a sufficiently accurate representation of the impulse response of line echo path 144.
  • X register 212 stores as many samples of x(k) as are necessary for a sufficiently accurate representation of the impulse response of acoustic echo path 132.
  • Coefficient register 208 is preferably of sufficient size to store the filter coefficients for background filter 156, or background filter 166, but not both simultaneously. As should be clear, a smaller coefficient register would not permit background filters 156 and 166 to operate properly, whereas a larger coefficient register would waste memory space.
  • X register 212 stores N samples x(k), x(k-1), ... x(k-N-1) of receive input signal x(k), with each sample having a length of M bits
  • U register 202 stores N/2 samples u(k), u(k-1), ... u(k-N/2-1) of send input signal u(k), with each sample having a length of M bits.
  • background filter 166 would employ N single-precision filter coefficients, each having a length of M bits, and background filter 156 would employ N/2 double-precision filter coefficients, each having a length of 2M bits. This would provide background filter 156 with roughly twice the accuracy of background filter 166.
  • coefficient register 208 would use N addressable memory locations, with each memory location having a length of M bits. Of course, the same-sized coefficient register would be preferred if background filters 156 and 166 each used N filter coefficients of M bits each, or N/2 filter coefficients of 2M bits each.
  • Switching coefficient register 208 between background filters 156 and 166 includes resetting coefficient register 208. As part of switching coefficient register 208 from U register 202 to X register 212, control unit 204 resets coefficient register 208 before allowing adaptation processor 210 to begin generating filter coefficients for background filter 166. Likewise, as part of switching coefficient register 208 from X register 212 to U register 202, control unit 204 resets coefficient register 208 before allowing adaptation processor 216 to begin generating filter coefficients for background filter 156. The resetting prevents swapping filter coefficients between background filters 156 and 166.
  • Coefficient register 208 may be reset merely by clearing it. However, it is possible that a large portion of line echo path 144, or acoustic echo path 132, or both, is determined by telephone set 100, and therefore is known. Accordingly, echo paths 132 and/or 144 may be reasonably approximated by resetting coefficient register 208 to predetermined values under appropriate circumstances. For example, if the echo canceller being switched to coefficient register 208 has a cascade arrangement, such as in Yatsuzuka et al., supra, then coefficient register 208 may be reset to predetermined filter coefficients when the foreground filter coefficients are cleared. Alternatively, if the echo canceller being switched to coefficient register 208 has a parallel arrangement, such as in Ochiai et al, supra, then coefficient register 208 may be reset to predetermined filter coefficients regardless of the foreground filter coefficients.
  • the first and second echo cancellers have each consisted of a two filters.
  • the first and second echo cancellers may each employ a single filter, or three filters, or more.
  • the present invention is also well suited for sharing a coefficient memory, or several coefficient memories, between filters in various configurations such as parallel or cascade arrangements.
  • the filters may be designed solely for canceling acoustic echos, line echos, or any combination of acoustic and line echos.

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
  • Filters That Use Time-Delay Elements (AREA)
  • Soundproofing, Sound Blocking, And Sound Damping (AREA)
EP96305918A 1995-08-25 1996-08-14 Echokompensatorsystem mit gemeinschaftlichem Koeffizientsspeicher Withdrawn EP0760574A2 (de)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US08/519,500 US5663955A (en) 1995-08-25 1995-08-25 Echo canceller system with shared coefficient memory
US519500 1995-08-25

Publications (1)

Publication Number Publication Date
EP0760574A2 true EP0760574A2 (de) 1997-03-05

Family

ID=24068570

Family Applications (1)

Application Number Title Priority Date Filing Date
EP96305918A Withdrawn EP0760574A2 (de) 1995-08-25 1996-08-14 Echokompensatorsystem mit gemeinschaftlichem Koeffizientsspeicher

Country Status (4)

Country Link
US (1) US5663955A (de)
EP (1) EP0760574A2 (de)
JP (1) JP3159922B2 (de)
KR (1) KR100559752B1 (de)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0918430A2 (de) 1997-10-31 1999-05-26 Lucent Technologies Inc. Echokompensatorn einem Netz für Datenanwendungen

Families Citing this family (32)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0843934B1 (de) * 1996-05-31 2007-11-14 Koninklijke Philips Electronics N.V. Einrichtung zur unterdrückung einer störenden komponente eines eingangssignals
US5937009A (en) * 1996-06-28 1999-08-10 Wong; Kon Max Sub-band echo canceller using optimum wavelet packets and cross-band cancellation
JP3396393B2 (ja) * 1997-04-30 2003-04-14 沖電気工業株式会社 エコー・雑音成分除去装置
US6185195B1 (en) * 1997-05-16 2001-02-06 Qualcomm Incorporated Methods for preventing and detecting message collisions in a half-duplex communication system
US6108412A (en) * 1997-10-07 2000-08-22 Nortel Networks Corporation Adaptive echo cancelling system for telephony applications
US6256383B1 (en) * 1997-11-07 2001-07-03 Legerity, Inc. IIR filter of adaptive balance circuit for long tail echo cancellation
US6181793B1 (en) * 1997-11-14 2001-01-30 Tellabs Operations, Inc. Echo canceller employing dual-H architecture having improved coefficient transfer
US6031908A (en) * 1997-11-14 2000-02-29 Tellabs Operations, Inc. Echo canceller employing dual-H architecture having variable adaptive gain settings
US6028929A (en) * 1997-11-14 2000-02-22 Tellabs Operations, Inc. Echo canceller employing dual-H architecture having improved non-linear echo path detection
US6240180B1 (en) * 1997-11-14 2001-05-29 Tellabs Operations, Inc. Echo canceller employing dual-H architecture having split adaptive gain settings
US6256384B1 (en) 1997-12-02 2001-07-03 Telefonaktiebolaget Lm Ericsson (Publ) Method and apparatus for cancelling echo originating from a mobile terminal
US5987098A (en) * 1997-12-23 1999-11-16 Nortel Networks Corporation Method and system for sparing echo cancellers
US6381272B1 (en) * 1998-03-24 2002-04-30 Texas Instruments Incorporated Multi-channel adaptive filtering
US6381238B1 (en) * 1998-07-31 2002-04-30 Sonus Networks Apparatus and method for a telephony gateway
US7035396B1 (en) * 1999-01-22 2006-04-25 Agere Systems Inc. Configurable echo canceller
US6694020B1 (en) * 1999-09-14 2004-02-17 Agere Systems, Inc. Frequency domain stereophonic acoustic echo canceller utilizing non-linear transformations
KR100436480B1 (ko) * 2000-03-08 2004-06-24 엘지전자 주식회사 교환 시스템에서 적응형 반향 제거기의 테이블 관리 방법
CN1243416C (zh) * 2000-03-27 2006-02-22 朗迅科技公司 用自适应判决门限来完成通话重叠检测的方法和装置
US6434235B1 (en) 2000-08-01 2002-08-13 Lucent Technologies Inc. Acoustic echo canceler
GB2389286A (en) 2002-05-28 2003-12-03 Mitel Knowledge Corp Echo cancellation
WO2004107724A2 (en) * 2003-05-27 2004-12-09 Koninklijke Philips Electronics N.V. Loudspeaker-microphone system with echo cancellation system and method for echo cancellation
US7349353B2 (en) * 2003-12-04 2008-03-25 Intel Corporation Techniques to reduce echo
EP1796281B1 (de) * 2005-12-09 2015-08-12 Mitel Networks Corporation Echounterdrücker
WO2008112538A1 (en) * 2007-03-09 2008-09-18 Fortemedia, Inc. Acoustic echo cancellation system
JP5160113B2 (ja) * 2007-03-22 2013-03-13 株式会社日立メディコ 磁気共鳴イメージング装置
US8050398B1 (en) * 2007-10-31 2011-11-01 Clearone Communications, Inc. Adaptive conferencing pod sidetone compensator connecting to a telephonic device having intermittent sidetone
SE533956C2 (sv) * 2009-07-20 2011-03-15 Limes Audio Ab Enhet och metod för styrning av restekodämpning
US20120140918A1 (en) * 2010-12-06 2012-06-07 Pagebites, Inc. System and method for echo reduction in audio and video telecommunications over a network
US9613634B2 (en) * 2014-06-19 2017-04-04 Yang Gao Control of acoustic echo canceller adaptive filter for speech enhancement
EP3274992B1 (de) 2015-03-27 2020-11-04 Dolby Laboratories Licensing Corporation Adaptive audiofilterung
US9479650B1 (en) 2015-05-04 2016-10-25 Captioncall, Llc Methods and devices for updating filter coefficients during echo cancellation
EP3667662B1 (de) * 2018-12-12 2022-08-10 Panasonic Intellectual Property Corporation of America Vorrichtung zur unterdrückung von akustischem echo, verfahren zur unterdrückung von akustischem echo und programm zur unterdrückung von akustischem echo

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2469044A1 (fr) * 1979-08-30 1981-05-08 Thomson Csf Mat Tel Procede de traitement en temps partage de signaux numeriques et application a un annuleur d'echo autoadaptatif multiplexe
FR2517906A1 (fr) * 1981-12-03 1983-06-10 Centre Nat Rech Scient Annulateur d'echo a commande automatique de gain pour systemes de transmission
US4628156A (en) * 1982-12-27 1986-12-09 International Business Machines Corporation Canceller trained echo suppressor
US4629829A (en) * 1984-12-14 1986-12-16 Motorola, Inc. Full duplex speakerphone for radio and landline telephones
JPS63236422A (ja) * 1987-03-25 1988-10-03 Hitachi Ltd 二線全二重モデム用エコ−キヤンセラ
NL8701633A (nl) * 1987-07-10 1989-02-01 Philips Nv Digitale echocompensator.
US4912758A (en) * 1988-10-26 1990-03-27 International Business Machines Corporation Full-duplex digital speakerphone
US5146494A (en) * 1989-07-31 1992-09-08 At&T Bell Laboratories Overlapping look-up-and-add echo canceller requiring a smaller memory size
NL9002790A (nl) * 1990-12-18 1992-07-16 Philips Nv Echocompensator met verbeterde dubbelspraak detectie.
JP2836277B2 (ja) * 1991-03-14 1998-12-14 国際電信電話株式会社 エコーキャンセル装置
AU659842B2 (en) * 1991-06-13 1995-06-01 Nec Corporation Method and arrangement of echo elimination in digital telecommunications system
JP2792311B2 (ja) * 1992-01-31 1998-09-03 日本電気株式会社 多チャンネルエコー除去方法および装置
US5353348A (en) * 1993-05-14 1994-10-04 Jrc International, Inc. Double echo cancelling system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0918430A2 (de) 1997-10-31 1999-05-26 Lucent Technologies Inc. Echokompensatorn einem Netz für Datenanwendungen
EP0918430A3 (de) * 1997-10-31 2004-12-08 Lucent Technologies Inc. Echokompensatorn einem Netz für Datenanwendungen

Also Published As

Publication number Publication date
JPH09130309A (ja) 1997-05-16
KR100559752B1 (ko) 2006-10-11
US5663955A (en) 1997-09-02
JP3159922B2 (ja) 2001-04-23

Similar Documents

Publication Publication Date Title
US5663955A (en) Echo canceller system with shared coefficient memory
US5664011A (en) Echo canceller with adaptive and non-adaptive filters
US6947549B2 (en) Echo canceller
US4757527A (en) Echo canceller
US5796819A (en) Echo canceller for non-linear circuits
US5668794A (en) Variable gain echo suppressor
US20090028354A1 (en) Echo Canceller Employing Dual-H Architecture Having Split Adaptive Gain Settings
US5764753A (en) Half-duplex controller
US5852661A (en) Adaptive echo cancellation used with echo suppression to reduce short and long duration echoes
US20060115077A1 (en) Echo canceller employing dual-H architecture having variable adaptive gain settings
US5247512A (en) Echo canceller
EP1264418B1 (de) Verfahren und vorrichtung für stereoechounterdrückung in einen voip-kommunikationsystemen
US6181793B1 (en) Echo canceller employing dual-H architecture having improved coefficient transfer
US5875246A (en) Distributed audio signal processing in a network experiencing transmission delay
JP3385221B2 (ja) エコーキャンセラ
US6975722B2 (en) Dual echo canceller with fast reflector identification
Makino et al. Echo control in telecommunications
JPH07303073A (ja) ノイズ低減回路およびこの回路を用いたエコーキャンセラ
JPH0465923A (ja) 音声信号処理装置とその方法
CA2592930A1 (en) Echo canceller employing dual-h architecture having variable adaptive gain settings

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): DE ES FR GB IT SE

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20020301